SCIENCE ADVANCES | RESEARCH ARTICLE

CELL BIOLOGY Copyright © 2018

The Authors, some

Tensile forces drive a reversible fibroblast-to- rights reserved;

exclusive licensee
myofibroblast transition during tissue growth in American Association
for the Advancement
engineered clefts of Science. No claim to
original U.S. Government
Works. Distributed
Philip Kollmannsberger,1,2*† Cécile M. Bidan,2,3 John W. C. Dunlop,2‡ Peter Fratzl,2 Viola Vogel1* under a Creative
Commons Attribution
Myofibroblasts orchestrate wound healing processes, and if they remain activated, they drive disease progres- NonCommercial
sion such as fibrosis and cancer. Besides growth factor signaling, the local extracellular matrix (ECM) and its License 4.0 (CC BY-NC).
mechanical properties are central regulators of these processes. It remains unknown whether transforming
growth factor–b (TGF-b) and tensile forces work synergistically in up-regulating the transition of fibroblasts into
myofibroblasts and whether myofibroblasts undergo apoptosis or become deactivated by other means once
tissue homeostasis is reached. We used three-dimensional microtissues grown in vitro from fibroblasts in
macroscopically engineered clefts for several weeks and found that fibroblasts transitioned into myofibroblasts
at the highly tensed growth front as the microtissue progressively closed the cleft, in analogy to closing a

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

wound site. Proliferation was up-regulated at the growth front, and new highly stretched fibronectin fibers were
deposited, as revealed by fibronectin fluorescence resonance energy transfer probes. As the tissue was grow-
ing, the ECM underneath matured into a collagen-rich tissue containing mostly fibroblasts instead of myofibro-
blasts, and the fibronectin fibers were under reduced tension. This correlated with a progressive rounding of
cells from the growth front inward, with decreased a–smooth muscle actin expression, YAP nuclear trans-
location, and cell proliferation. Together, this suggests that the myofibroblast phenotype is stabilized at the
growth front by tensile forces, even in the absence of endogenously supplemented TGF-b, and reverts into a
quiescent fibroblast phenotype already 10 mm behind the growth front, thus giving rise to a myofibroblast-to-
fibroblast transition. This is the hallmark of reaching prohealing homeostasis.

INTRODUCTION mately resulting in a loss of organ function (3, 6, 8, 11–13). Successful

Throughout our lifetime, most of our tissues remodel and regrow to treatment of these pathologies, as well as proper healing and remodeling
adapt to changing requirements and to regenerate from damage or dis- after injury or implantation, requires that tissue homeostasis be reestab-
ease. In wound repair, myofibroblasts are responsible for wound lished. This major clinical challenge can only be addressed by a better
closure. Their transition from tissue-resident fibroblasts or other mes- understanding of how biochemical and physical cues instruct fibroblasts
enchymal precursors (1) is promoted by soluble factors such as to proliferate and become myofibroblasts, contract the wound site, and
transforming growth factor–b1 (TGF-b1) and by mechanical cues deposit new ECM. Moreover, the repaired tissue must return to a quies-
(2–6). The fibroblast-to-myofibroblast transition is described as a cent, homeostatic state to prevent fibrotic states.
multifactorial process that involves enhanced force generation with en- In a wound site, many factors interplay, such as cell-generated forces
larged focal adhesions and incorporation of a–smooth muscle actin to contract a wound, the rigidity and composition of the ECM of the
(a-SMA) into stress fibers (7, 8). Furthermore, increased contraction injured tissue that is interwoven with the fibrin gel from the blood clot,
of extracellular matrix (ECM) fibers can uncage TGF-b from its latent and soluble factors released by blood cells entrapped in the blood clot
complex and release it into the medium (9). However, little is known, at and by invading cells, including fibroblasts and macrophages. Disen-
the organ level, about the fate of myofibroblasts after successful wound tangling this mechanically driven growth factor–cell-matrix recipro-
closure—whether they undergo apoptosis, become senescent, or revert city to understand how forces and ECM properties drive myofibroblast
to a fibroblast phenotype. Although washing away or blocking the differentiation and vice versa is difficult in a real wound model because
action of TGF-b1 in myofibroblast two-dimensional (2D) cell cultures too many factors contribute, and in 2D cell culture models due to their
is sufficient to revert the phenotype in vitro, until recently, the revers- inherent limitations. The native ECM deposited by cells on flat culture
ibility of the myofibroblast phenotype has not been demonstrated in substrates is limited to a thickness of 20 to 30 mm and remains randomly
animals [for review, see Van De Water et al. (6)], with one exception organized under static conditions (14), whereas 3D fibrous scaffolds,
in a rat liver fibrosis model (10). Pathological and persistent myofibro- such as collagen gels or matrigels, have a much higher fiber density than
blast activation can lead to hypertrophic scars or tissue fibrosis, ulti- native tissue and lack much of the biochemical complexity and remod-
eling dynamics of native tissue ECM. Although most drugs are still
1
Laboratory of Applied Mechanobiology, Institute of Translational Medicine, Department screened and tested in 2D cell culture, it is well established that the drug
of Health Science and Technology, ETH (Eidgenössische Technische Hochschule) Zurich, dose response (15) and radioresponse (16) depend on the 3D micro-
Zurich, Switzerland. 2Department of Biomaterials, Max Planck Institute of Colloids and environment that the cells experience. Moreover, the long-term effects
Interfaces, Golm, Germany. 3Université Grenoble Alpes, CNRS, Laboratoire Interdisciplinaire
de Physique, 38000 Grenoble, France.
of tissue maturation on myofibroblast differentiation (17) are poorly
*Corresponding author. Email: [email protected] (V.V.); philip.kollmannsberger@uni- understood, because existing in vitro approaches are often limited to
wuerzburg.de (P.K.) a few hours or days.
†Present address: Center for Computational and Theoretical Biology, University of Here, we studied the progressively changing relationship between
Würzburg, Würzburg, Germany.
‡Present address: FB (Fachbereich) Chemistry and Physics of Materials, Paris Lodron cellular forces, tissue maturation, and myofibroblast differentiation over
University of Salzburg, Salzburg, Austria. several weeks in an in vitro 3D microtissue model (18) that addresses

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 1 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE

several of these limitations. In this assay, de novo tissue was grown from dium at day 18 with EdU (5-ethynyl-2′–deoxyuridine), which incorpo-
fibroblasts, and the cells assembled their own ECM as they were grad- rates only into newly synthesized DNA (Fig. 2). Although EdU-positive
ually closing millimeter-sized concave scaffold clefts. New tissue nuclei appear throughout the tissue, the density of EdU-positive cells is
emerged at the growth front (19), whereas increasing tissue depth highest in the growth front (Fig. 2, A to C, fig. S2, and movie S2). This
corresponds to increasing tissue maturity as transient cellular forces could result either from a higher proliferation rate or from a higher over-
are gradually converted into a permanent tissue prestress (19). To ad- all cell density in the growth front. Therefore, we performed a quanti-
dress the challenge outlined above, we advanced the previous assay (19) tative analysis of proliferation by determining the ratio of EdU-positive
and grew normal human dermal fibroblast (NHDF) microtissues in the to the total number of cells as a function of tissue depth and hence
corners of polydimethylsiloxane (PDMS) scaffolds with rectangular tissue age. Our results show that relative EdU intensity, and hence cell
holes. We developed automated image analysis tools to quantify fibro- proliferation, decreased with tissue depth (Fig. 2E). EdU incorporation
nectin (FN) fiber stretch, proliferation, and myofibroblast differentiation was always higher at the growth front compared to the interior, with no
as a function of tissue age. With this approach, we aim to improve our significant differences between conditions, that is, control or treatment
understanding of how tissue maturation and tension-guided cell matrix with blebbistatin or TGF-b1 (Fig. 2F). The higher EdU incorporation at
reciprocity orchestrate the fibroblast-to-myofibroblast transition and the growth front correlated with a flatter and more polarized appearance
myofibroblast activation, as well as the transition back to a homeostatic, of the cell nuclei compared to a more isotropic nuclear shape in the
quiescent state. Our ultimate aim is to apply the insights derived from tissue interior, which is known to mediate the cellular response to physi-
these control principles to develop novel approaches to promote healing cal cues (21, 22). Although there was no significant difference in prolif-
and remodeling in the context of regenerative therapies. eration between conditions of different cell contractility, cell density was

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

significantly increased in TGF-b1–treated samples (Fig. 2, G and H),
suggesting that the overall proliferation during the entire growth period
RESULTS was higher.
Primary human dermal fibroblasts grow de novo
microtissues progressively closing engineered clefts Newly deposited FN fibers are more stretched at the growth
Microfabricated PDMS scaffolds with macroscopic square-shaped clefts front compared to FN fibers in the tissue interior
were attached to the bottom of passivated cell culture wells using spacers Because FN fiber tension was shown in 2D cell culture to be up-regulated
(Fig. 1A and fig. S5). After seeding, fibroblasts adhered to the FN–cross- with cytoskeletal tension (23, 24), we added fluorescence resonance
linked scaffold surfaces, but not to the well bottom, and started to pro- energy transfer (FRET)–labeled FN (5 mg/ml) together with a ninefold
liferate until a dense monolayer had formed on all sides, including on excess of unlabeled FN to the medium of growing tissues to prevent in-
the vertical cleft walls. After typically 3 to 4 days, the corners of the rect- termolecular FRET, and allowed cells to incorporate the FN-FRET
angular clefts started to fill with a thick 3D tissue (Fig. 1B). During the probe into their ECM for the last 24 hours of growth. Thus, we only
subsequent 2 to 3 weeks of culture, the tissues grew larger until they visualize the FN fibers that got assembled in the last 24 hours (14). To be
merged, and only a round opening remained in the center of the square- sensitive to a large range of conformational changes, we labeled dimeric
shaped scaffold clefts. This is in agreement with earlier tissue growth studies human plasma FN with acceptors on the native cysteins in modules
performed in macroscopic scaffolds with mouse preosteoblasts (18) and FNIII7 and FNIII15, respectively, and randomly with donors as previously
human mesenchymal stromal cells (20). To characterize the architecture described (23–25). The batch exploited here had a donor-to-acceptor
of the de novo formed microtissues, we fixed the tissues after 3 weeks in ratio of 2.5. After this incubation period, tissues were fixed, and 3D image
culture and stained them for actin, collagen-I (Col-I), and nuclei. We found stacks of the FN matrix in the tissues were recorded by exciting the do-
that both FN and Col-I were present throughout the tissue, consistent with nor fluorophores and recording the emission from the donor and the
previous observations in osteoblasts (Fig. 1E). Actin, FN, and Col-I were all acceptor fluorophores separately (Fig. 3A). As the efficiency of FRET
highly polarized and aligned at the tissue-medium interface but appeared decreases with increasing distance between donors and acceptors, a
less organized in the bulk of the more mature tissue. low intensity ratio between acceptor and donor means that FN fibers
are on average more stretched compared to a high intensity ratio (26).
Cytoskeletal tension up-regulates the volumetric growth We found that the FRET intensity ratio between acceptors and donors
rate of the tissue declined as FN fibers got progressively more stretched (Fig. 3, B to E,
To investigate the role of cellular traction forces in the de novo tissue and movie S3) in response to increased cytoskeletal tension, in agree-
growth, we down- or up-regulated the cytoskeletal contractility through ment with previous observations (23, 24, 26).
the supplementation of the medium by blebbistatin (10 mM) or TGF- To quantify differences in FN fiber tension between the tissue
b1 (1 ng/ml), respectively (Fig. 1I). Under control conditions, tissue growth front and interior, and hence as a function of tissue age, we
radius (see Materials and Methods) was 344 ± 13 mm, whereas under calculated the 3D Euclidean distance map of the tissue volume from
conditions of increased and reduced cytoskeletal tension, tissue radius the growth front. As a result, each voxel in the tissue gets assigned its
increased to 428 ± 16 mm and decreased to 298 ± 9 mm, respectively Euclidean distance to the tissue surface. We found that FN-FRET grad-
(Fig. 1, F to I, and movie S1). This shows that the volume of the tissue ually increased with the distance from the growth front, suggesting a
formed by NHDFs in rectangular scaffold pores as measured by the decrease of FN fiber stretch with tissue age (Fig. 3, F to H). Because this
radius depends on cytoskeletal tension. method only detects FN fibers assembled during the last 24 hours of
exposure to FN-FRET, the results show that cells in the growth front
The tissue-medium interface is high in proliferating cells and assemble a more stretched FN fiber network compared to cells in the
defines the growth front more mature interior (Fig. 3G). Regions of high FN fiber stretch coin-
To ask whether tissue volume increases mainly by cell proliferation in cide with regions where F-actin, FN, and Col-I ECM appear highly
the growth front, or within the entire tissue, we supplemented the me- aligned with the growth front, as described above.

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 2 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE

A B C E

FN
Col-I
D Nuclei
µm
0
50

100 µm

x
F Control G Blebbistatin H TGF-β I
y

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

Actin

r
Nuclei

z
-

x
Fig. 1. De novo microtissue growth rate increased with cytoskeletal tension. (A) Schematic of the microfabrication process used to produce thick 3D PDMS
substrates with controlled macroscopic geometry. Detailed illustrations of the fabrication process are shown in fig. S5. (B) Illustration of 3D tissue growing in the
corners of rectangular substrate pores. (C and D) Illustration of surface and interior regions and tissue dimensions. (E) Maximum intensity projection of the resulting
3D tissue containing cells (blue) and ECM (red, FN; green, Col-I; blue, nuclei). The tissue is rich in both FN and Col-I fibers. (F to H) Single slices through tissues (green,
actin; blue, nuclei) at the indicated depth in xy and xz directions, under normal conditions (left) and under conditions of inhibited (middle) and elevated (right)
cytoskeletal tension in the presence of 10 mM blebbistatin or TGF-b1 (1 ng/ml), respectively. The individual channels and an animated z fly-through are shown in
fig. S1 and movie S1, respectively. (I) Quantification of tissue radius (distance between the scaffold corner and the growth front at 40 mm of z depth) after 19 days
of growth under normal conditions and in the presence of 10 mM blebbistatin or TGF-b1 (1 ng/ml), respectively. Increased cytoskeletal tension results in increased tissue
volume. Scale bars, 50 mm. Data points indicate the tissue radius for individual tissues.

Cytoskeletal tension regulates FN fiber tension in the ECM a-SMA–positive cells were found to reside in the tissue interior (Fig. 4,
at the growth front A to C, fig. S4, and movie S4). Quantification of the normalized a-SMA
We next determined whether the molecular stretch in the newly de- signal density as a function of the distance to the growth front reveals a
posited FN fibrils in the growth front is sensitive to cytoskeletal tension steep decline of a-SMA–expressing cells over the very first 10 mm be-
and how this might correlate with the tissue growth rate. When blebbistatin hind the growth front.
was present at low concentrations (10 mM) during the entire growth period
to allow for cell proliferation while reducing cytoskeletal tension, the FN TGF-b1 or blebbistatin supplementations tune a-SMA
fibrils at the growth front were on average less extended compared to expression levels but do not affect the overall phenomenon
control conditions (Fig. 3F). In contrast and when cytoskeletal tension Because the expression of a-SMA was associated with TGF-b–induced
was increased by supplementing the medium with TGF-b at a concen- myofibroblast differentiation (8), we next determined whether the ob-
tration of 1 ng/ml during the entire growth period, FN fibrils were more servation that a-SMA is mostly expressed within the first 10 mm of the
stretched compared to the control. In all cases, FN was significantly growth front depends on TGF-b stimulation by adding exogenous
more extended close to the growth front compared to the tissue interior. TGF-b1 or blebbistatin to increase and inhibit cytoskeletal tension
and the associated pathways, respectively. The expression of a-SMA at
a-SMA is expressed mostly at the growth front, even the growth front was significantly increased under the presence of TGF-
without TGF-b supplementation b1 (1 ng/ml), whereas the presence of 10 mM blebbistatin had a slight but
Because a-SMA expression is characteristic for the more contractile not statistically significant decreasing effect on a-SMA expression (Fig. 4,
myofibroblasts, but not fibroblasts, the microtissues were grown with- D to F). In all cases, the level of a-SMA was still significantly higher at
out TGF-b supplementation and stained for a-SMA (red) after fixation the growth front compared to the tissue interior. The fact that the pres-
and counterstained for actin with phalloidin (green). Significantly, ence of supplemented TGF-b1 cannot convert the resident fibroblasts in
the growth front is enriched in a-SMA, whereas only a small number of the tissue interior into a myofibroblast phenotype suggests that the

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 3 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE
x
A Control D
y

E F

(surface/interior)
z

x
B Blebbistatin

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

G H
Nuclear volume fraction

Nuclear volume fraction

C TGF-

Fig. 2. Spatial mapping of EdU incorporation reveals that cells preferentially proliferate in the surface growth front. (A to C) Single slices through tissues at the
indicated z depth showing EdU signals from proliferating nuclei (green) and Hoechst counterstain of all nuclei (red) for all three conditions in xy and xz directions. Scale
bars, 50 mm. The individual channels and an animated z fly-through are shown in fig. S2 and movie S2, respectively. (D) Time course of proliferation experiments. Tissues
were cultured for 19 days under normal conditions and in the presence of 10 mM blebbistatin or TGF-b1 (1 ng/ml), respectively. (E) Percentage of EdU-positive volume
compared to total nuclear volume as a function of tissue depth and age. (F) Ratio of EdU-positive cells in growth front compared to interior. (G) Nuclear volume fraction
as a measure for cell density as a function of tissue depth and age. (H) Mean nuclear volume fraction for all three conditions. Error bars indicate SEM. Data points in (F)
and (H) indicate ratios and fractions for individual tissues.

stimulation of these microtissues by exogenous TGF-b1 cannot override interior compared to the control and the blebbistatin-treated tissues
the stimulation imposed by the reduced tensile forces compared to the (10 mM) (Fig. 4, J to L).
growth front, as well as by the ECM microenvironment.

YAP nuclear localization is highest at the growth front DISCUSSION

Because the nuclear translocation of YAP is up-regulated in response to Although fibroblasts in 2D cell culture can be stimulated with TGF-b1
mechanical signals and promotes cell proliferation (27, 28), we next in- to transition into myofibroblasts and revert back to the fibroblast pheno-
vestigated the translocation of the transcription factor YAP/TAZ from type when TGF-b1 is washed away or by blocking the action of TGF-
cytoplasm to nucleus as a function of the cell distance to the growth b1, the reversibility of the myofibroblast phenotype has not been dem-
front. For the 3D distance analysis from the growth front, we used onstrated in animals until recently [for review, see Van De Water et al.
4′,6-diamidino-2-phenylindole (DAPI)–stained nuclei as masks to dis- (6)], with one exception in a rat liver fibrosis model (10). Even with-
tinguish the nuclear and cytoplasmic signals (Fig. 4, G to I, and movie S5). out TGF-b1 supplementation, we found a high density of myofibro-
We found that the percentage of nuclear YAP signal intensity decreased blasts in the highly tensed growth front of microtissues that were
from nearly 55% at the growth front down to about 35% within the tissue grown in engineered clefts, whereas the tissue interior that filled the
interior, indicating a significantly larger degree of nuclear localization and corners contained mostly quiescent fibroblasts (Figs. 3 and 4). Further-
therefore activity of YAP at the growth front (Fig. 4, J to L). Upon TGF- more, cell proliferation (Fig. 2), expression of contractile a-SMA fibers
b1 supplementation (1 ng/ml), the YAP nuclear localization again de- (Fig. 4, A to F), and nuclear localization of YAP (Fig. 4, G to L) were
clined rapidly within the first 10 mm, and was actually lower in the tissue all significantly up-regulated along the growth front compared to the

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 4 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE

A F G

B in GdnHCI solution

x
Control Blebbistatin TGF-
y C D E

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

z

Fig. 3. FN fiber tension is highest in the growth front and decays as the tissue matures. (A) Time course showing the late supplementation of the medium by the
FN-FRET probes. (B) Histograms of FN-FRET ratios of FN fibers within the ECM for three different medium supplementations. The FN-FRET probe was added to the
medium 24 hours before fixing the cells. The FN-FRET ratios in solution for the same batch are also given in different concentrations of the denaturant GdnHCl (gray
lines). (C to E) Single slices through tissues at the indicated z depth showing the FN-FRET ratio for all three conditions. Scale bars, 50 mm. Notice how overall tissue
volume depends on up-regulation (TGF-b1, 1 ng/ml) or down-regulation (10 mM blebbistatin) of cell contractility. An animated z fly-through is shown in movie S3. (F) Median
FRET ratio is significantly different between all three conditions. (G) Median FN-FRET ratio was always higher in the interior compared to the growth front, with no significant
differences between the three conditions. (H) FN-FRET ratio as a function of distance to the growth front or, equivalently, tissue maturity, for all three conditions. Tissues were
cultured for 19 days. Error bars indicate SEM. Data points in (F) and (G) indicate median values and ratios for individual tissues.

interior. FN-FRET probes revealed that the newly deposited FN fibers contractile forces acting on the scaffold walls at an angle result in a lift-
were significantly more stretched along the growth front compared to off, as observed previously (18, 19, 29), resulting in a bridging of the
the tissue interior that gradually filled up the corners and were aligned opposing scaffold walls by a highly contractile cell sheet while gradu-
parallel to the actin cable (Fig. 3). In Fig. 5, these differences in cell phe- ally filling in the scaffold corners. Tensile force acting on the cells is
notype and matrix stretch that define “tissue maturation” are graphi- highest in the growth front, and we propose here that they drive the
cally summarized. Supplementation of TGF-b1 to the medium only transition of fibroblasts into myofibroblasts, and thus the fibroblast-to-
enhanced the phenomena seen without TGF-b supplementation but myofibroblast transition, even in the absence of TGF-b supplemen-
did not change it fundamentally: TGF-b1 supplementation increased tation. Enhancing cell contractility up-regulates the tissue growth rate,
the tissue growth rate (Fig. 1), promoted higher stretch of FN fibrils as shown here by supplementing the medium with TGF-b1 (Fig. 1),
in the growth front and in the interior (Fig. 3), increased a-SMA expres- and the same effect can be induced by decreasing the cleft angle (18).
sion along the growth front and its thickness (Fig. 4), and reduced nuclear This interpretation is in agreement with in vivo data because cyto-
localization of YAP in the tissue interior (Fig. 4). In contrast, inhibition of skeletal tension was reported to up-regulate myosin dynamics during
myosin-II with blebbistatin resulted in lower tissue volume (Fig. 1) and Drosophila embryonic wound repair (4), whereby myosin activity is
lower FN fiber extension in ECM (Fig. 3) and had little effect on YAP regulated by RhoA/Rock. Also in agreement with the data, we found
nuclear translocation compared to the control (Fig. 4). previously that FN fiber tension in early ECM is up-regulated in a
These findings, as summarized in Fig. 5, suggest the following suc- RhoA-dependent manner (25).
cession of events during tissue growth: First and directly after seeding, Second, the myofibroblasts at the growth front continue to assemble
fibroblasts migrate and proliferate homogeneously on the scaffold sur- a highly stretched FN matrix that, as the tissue growth front advances,
face until they colonize the entire surface. Within the first 3 days after matures underneath and gets interlaced with Col-I fibers (19). The
seeding, a dense and contractile cell monolayer forms on all surfaces of growth front seems to bear the major load acting on the opposing scaf-
the scaffold with a highly stretched FN-rich ECM that initially does not fold walls because the cells in the maturing tissue have a more rounded
grow thicker than 20 to 30 mm, as observed on other planar culture sub- cell shape. The FN matrix deposited and remodeled by these cells in
strates before (14). As time progresses (typically after 4 to 5 days), the the tissue interior was less stretched, and cells adopted a quiescent,

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 5 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE
x x
Control Control
A D J
y y G

Actin Nuclei
aSMA YAP

z z

x x
Blebbistatin Blebbistatin

B H

E F K L

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

TGF-
TGF-
C I

Fig. 4. a-SMA intensity and YAP nuclear localization were highest in the growth front. (A to C) Single slices through tissues at the indicated locations for all three
conditions. F-actin is shown in green, and a-SMA fibers are shown in red. Individual channels are shown in fig. S3, and an animated z fly-through in movie S4. (D) Ratio
of a-SMA signal density in fibers and in background as a function of tissue depth and age shows preferential a-SMA expression on the outermost layer within 10 mm
from the tissue growth front. Addition of TGF-b1 (1 ng/ml) results in an increase of a-SMA signal, whereas blebbistatin (10 mM) reduces it. (E) Average a-SMA expression
on surface and (F) ratio of a-SMA expression between growth front and interior. (G to I) Single slices through tissues at the indicated locations showing YAP signal
(cyan) overlaid with nuclei (red) for all three conditions. Individual channels are shown in fig. S4, and an animated z fly-through is shown in movie S5. (J) Ratio of YAP
signal in nuclei over total YAP signal shows a decrease from up to 60% nuclear localization on the tissue surface to nearly 25% deeper in the tissue. Addition of TGF-b
results in a larger difference (gradient) in nuclear localization between tissue growth front and interior. (K) YAP nuclear localization in the growth front and (L) ratio of
YAP nuclear localization between growth front and interior. Tissues were cultured for 19 days. Error bars indicate SEM. Data points in (E) and (F), and (K) and (L) indicate
ratios and percentages for individual tissues. Scale bars, 50 mm.

less proliferative phenotype with low a-SMA expression and cytosolic adding blebbistatin before the experiment resulted in a smaller nick angle,
localization of YAP. The presence of thicker and partially radially indicating the presence of both actively generated and permanently stored
oriented collagen fibers in the interior furthermore indicates maturation mechanical tension at the tissue edge.
of the tissue with increasing age (19). Because collagen rather than FN is Our data show that the activation of fibroblasts into contractile,
the major force-bearing component of mature ECM (30), these more proliferative myofibroblasts is transient and that myofibroblasts can
rigid collagen cables likely shield the FN fibers from cell-generated ten- undergo a myofibroblast-to-fibroblast transition in the presence of
sion as the FN and collagen meshworks are interwoven, or because the TGF-b (Fig. 4). Even without addition of exogenous TGF-b, we ob-
cells preferentially engage with the collagen integrin receptors. This, served expression of a-SMA in the growth region (Fig. 4), but not in
together with the more rounded and thus less contractile fibroblasts the interior at later time points. Because all cells were exposed to the
(Fig. 1), can account for the reduced FN fiber tension seen in the same growth medium, it is the cell-generated tissue microenvironment
maturing tissue interior. as modulated by cell-generated traction forces that drives the fibroblast-
Previous studies in collagen-based contractile 3D microtissues have myofibroblast-fibroblast transition. Future research has to reveal whether
shown a direct connection between tissue tension and geometry as defined this transition is driven by tension gradients alone or whether other
by the anchor points, evident by cell and collagen fiber alignment (31), but factors contribute as well. These factors could include switching the ex-
not FN fiber stretch (32), likely due to the presence of thick collagen posure of binding sites of integrin, growth factor, or cytokines on FN
bundles. Microsurgically induced wounds in these tissues healed within fibers by releasing fiber tension or an up-regulated presence of collagen
24 hours, with highly motile cells contracting the tissue and pulling new fibers and of other ECM components.
FN fibers to close the gap (33). Also, this agrees well with our observation Our results suggest that the cells, starting from a homogeneous
of increased mobility and FN fiber assembly at the growth front compared population of fibroblasts, transition into contractile, proliferative, ECM-
to the interior (movie S6). We performed laser cutting experiments in a synthesizing myofibroblasts in the growth front and revert back into
previous study (19), inspired by the developmental biology literature (34). quiescent fibroblasts at later time points in the more mature tissue
Cutting the actin cable-like growth front caused an instantaneous nick, interior. This is remarkable, because the fate of myofibroblasts after a
which “healed” within 24 hours. Reducing the active cellular tension by successful wound closure is still debated (3, 6, 10). Although washing

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 6 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE

myofibroblast markers, increased contractility, and disturbed ECM re-

Quiescent fibroblasts modeling, which are mediated by TGF-b signaling (3, 5, 9, 38). For a
HOMEOSTASIS wound to heal properly, the right balance of fibroblast and myofibro-
blast phenotype must be maintained, but strategies to efficiently control
Low on this balance are currently lacking (8). Here, we discovered that the abil-
Proliferation nsi
Te
ity of cells to generate contractile forces is the key parameter that con-
Cell ellipticity
FN fiber tension
trols the transient appearance of a myofibroblast-like phenotype and,
Tis FN/Collagen I ratio promoted by a proper ECM architecture and composition, its reversal
su
em YAP nuclear translocation back into quiescent fibroblasts. The steepness of the gradients in ECM
a α-smooth muscle actin
tu architecture and composition depends on, and can be controlled by, the
ra
tio High macroscopic geometry of a wound or scaffold (19, 33). Although our
n model system lacks the provisional fibrin matrix that fills the wound
space and gets remodeled or resorbed in a physiological scenario, the
ion

principles we discovered still apply: Contraction of a wound requires

Activated Growth
ns

healthy surrounding tissue to balance the contractile force. In addition

Te

myofibroblasts to using scaffolds as guiding structures for ECM organization during

tissue regeneration (39), the dynamic feedback process between cells
Fig. 5. The myofibroblast-to-fibroblast transition illustrating how the highly
and matrix acting during the growth process must be taken into account

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

tensed growth front rich in myofibroblasts drives tissue growth and how
myofibroblasts revert to quiescent fibroblasts as the tissue matures. As the
to induce proper healing. The same principles apply during embryogen-
tissue grows tensed by cell contractility, and new cells and ECM are added at the esis, where geometry-dependent patterns of tensile forces are increasingly
growth front in a layer-by-layer process, a gradient emerges from the growth surface recognized as cues for controlling cell fate and tissue structure (40).
toward the tissue interior. The arrow of tissue maturation denotes the emerging gra- Our findings thus have significant implications for understanding
dients in tissue properties over space and time from the surface to the core. In the and controlling biological processes of growth and regeneration in
growth front, cells proliferate, are well spread and elongated, and deposit and stretch the context of wound healing, tissue engineering, development, and
an early, 2D FN-rich ECM, whereas within the tissue interior, cells are more isotropic stem cells.
and embedded into a 3D Col-I–rich ECM. This gradient in ECM properties correlates
with a gradient in cell phenotype, from activated myofibroblasts in the growth front
toward a quiescent fibroblast phenotype that maintains tissue homeostasis in the interior.
MATERIALS AND METHODS
Substrate fabrication
away or blocking the action of TGF-b1 in myofibroblast cultures was Master fabrication
sufficient to revert the phenotype in vitro (6), it was only recently shown Master structures of tissue growth substrates were made using micro-
that not all myofibroblasts undergo apoptosis in vivo but that some scope projection lithography (41), summarized in fig. S5. This method
revert back to an inactive fibroblast phenotype in liver fibrosis (10, 35). allows for fast and easy fabrication of thick master structures at a frac-
We cannot completely rule out myofibroblast apoptosis in our micro- tion of the cost of standard photolithography and without the need
tissues as an alternative explanation, yet the absence of apoptotic cells for clean room equipment. With careful tuning of the exposure para-
on the growth front and the low proliferation rate in the tissue interior meters, resolutions down to few micrometers can be achieved, which
speak against this interpretation. is sufficient for the large-scale features used in this study. We used a
If the fibroblast-to-myofibroblast transition would be irreversible, variation of the method described by Kim et al. (42). A layer of poly-
then there should be a higher number of myofibroblast-like compared ethylene glycol diacrylate (PEG-DA; molecular weight, 700; Sigma-
to fibroblast-like cells, because proliferation is much higher at the Aldrich) mixed with a photocrosslinker (Irgacure 2959, BASF) at
growth front. This gradient in cell phenotype and ECM already emerges 2% (w/v) was sandwiched between a microscopy glass slide that had
at earlier time points (fig. S6). The only other explanation is migration of been activated with 3-(trimethoxysilyl)propyl methacrylate (Sigma-
nondifferentiated cells from outside the tissue into the interior. How- Aldrich) and a coverglass (#1, 24 mm × 60 mm) passivated using
ever, time lapse recording of the tissue dynamics (movie S6) shows a fluorosilane [trichloro(1H,1H,2H,2H-perfluorooctyl)silane, Sigma-
very dynamic growth surface but a rather static interior. This indicates Aldrich] in vapor phase. Two standard coverslips (#1, 22 mm ×
enhanced FN fiber assembly at the growth front but contradicts the 22 mm) were used as spacers, leading to a feature thickness of about
presence of enhanced cell migration in the interior. Moreover, the same 340 mm. First, a frame was polymerized around the feature area by
FN matrix assembled by the cells in the growth front is found a few days covering it with a piece of opaque tape and exposing it to the light of
later in the tissue interior (19). Therefore, a transient phenotypic change an ultraviolet (UV) lamp (365 nm). Subsequently, the tape was re-
is the only model that is consistent with our data. moved, and the master was mounted on an inverted fluorescence mi-
Developing a mechanistic understanding of whether myofibroblasts croscope (Zeiss Axiovert 200m, Zeiss) with the coverslip facing the
undergo apoptosis or revert back to a fibroblast phenotype has major objective, and a rectangular mask (2 mm × 2 mm) was inserted into
implications on how to treat progressive fibrotic diseases where learning the field stop slider of the fluorescence pathway. Using the automated
how to revert the tissue back to homeostasis is urgently needed (3, 6, 10, 35). shutter and motorized XY stage of the microscope, a regular array of
Upon tissue damage, myofibroblasts are responsible for wound repair rectangular features was polymerized by illuminating the master with
and closure, and excessive fibroproliferation is a common complication the DAPI channel of the microscope through a 5× objective (numerical
after injury (36). Aberrations from normal wound healing, resulting in aperture, 0.12) using an exposure time of 500 ms. After exposure, the
prolonged inflammation, scarring, and fibrosis, present a major clinical top coverslip was carefully removed, and the remaining prepolymer
challenge (37). These complications are associated with expression of was washed off using acetone.

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 7 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE

Microfabrication of PDMS scaffolds Abcam plc) at 1:100 and counterstained with Alexa Fluor 546 goat anti-
To create PDMS replicas, the master with the negative features was first mouse secondary antibody (A-11030, Thermo Fisher Scientific) at 1:100.
passivated using fluorosilane in vapor phase. A positive stamp was made, After washing with phosphate-buffered saline (PBS), primary and sec-
passivated, and used as a master for another negative structure in PDMS. ondary antibodies were applied for 30 min each except for Col-I primary
This secondary negative master was then passivated and used to print antibody, as mentioned above.
the final scaffold grids. For making the scaffolds, PDMS was mixed in Fabrication of the FN-FRET probe
the standard ratio of 10:1 and poured over the negative PDMS master, FN was isolated from human plasma, FRET-labeled, and characterized
slightly overfilling the structure to fill all gaps, and degassed at 50 bar for according to previously published protocols (24). In short, FN-DA was
30 min to remove trapped air bubbles. After precuring for 6 min at 80°C, double-labeled with Alexa Fluor 546 C-5 maleimide (A10258, Thermo
the excess PDMS was skimmed with a glass slide, and the remaining Fisher Scientific) as acceptor on free cysteines, and with Alexa Fluor 488
PDMS was cured for another 2 hours at 80°C. This proved to be the most succinimidyl ester (A20100, Thermo Fisher Scientific) as donor on ly-
reliable method to generate open PDMS grids without residual membranes. sines, and stored in PBS (0.5 mg/ml) at −80°C. The labeling ratio was
After removal from the master, the substrates were stored in ethanol. determined by UV absorption measurements as 3.9 acceptors and 8.8
Functionalization and passivation of the PDMS scaffolds donors per molecule, respectively. The FN-FRET probe was calibrated
To prepare the scaffolds for tissue growth experiments (as summarized as reported previously (24, 25) by measuring the FRET intensity ratios
in fig. S5), they were first cleaned and sterilized in 70% ethanol in an upon denaturation by different concentrations of GdnHCl in solution.
ultrasonic bath for 20 min. Subsequently, FN was covalently attached For FN-FRET experiments, FN-DA was mixed with unlabeled FN at a
to the surfaces of the scaffold using a heterobifunctional cross-linker ratio of 1:9 to prevent intermolecular energy transfer and added to the

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

(Sulfo-SANPAH, 1 mg/ml, Thermo Fisher Scientific), as described pre- culture medium at the specified time point at a concentration of 50 mg/ml.
viously (43). The procedure was carried out for each side separately to Proliferation assay
ensure homogeneous functionalization. Using optical glue NOA-61 EdU (Thermo Fisher Scientific) was added to the cell medium of grow-
(Norland Optical Adhesive 61, Norland) and small PDMS spacers, the ing tissues at 10 mM at day 18, where it was incorporated into newly
functionalized substrates were then attached to an ibidi-treated glass- synthesized DNA during the G1/S phase of the cell cycle. Cells were fixed
bottom dish (m-Dish, 35 mm, ibidi) that had previously been passivated after 16 hours, stained for EdU with Alexa Fluor 647 azide (A10277,
using PLL (poly-L-lysine)-g-PEG (0.1 mg/ml, SuSoS). This construction Thermo Fisher Scientific), and counterstained for chromatin using
restricted cell growth to the scaffold while allowing for imaging through Hoechst 33342 (H3570, Thermo Fisher Scientific) (Fig. 3A). The num-
the bottom coverglass. After assembly, the culture system was incubated ber of EdU-positive cells relative to the total number of cells then yielded
with medium at 37°C for at least 1 hour before seeding cells to equilibrate a readout for the probability of cells to be in the G1/S phase during the
the surface and allow air bubbles to escape from the pores. incubation period.

Microtissue growth experiments Microscopy and image analysis

Cell culture Confocal microscopy
Primary NHDFs (passage numbers 3 to 6, Lonza) were maintained in Tissues were imaged on an Olympus FV1000 confocal laser scanning
a–minimum essential medium supplemented with 10% fetal bovine microscope for FN-FRET using a 10× objective. FRET donor excitation
serum and 1% penicillin-streptomycin. For experiments, cells were was at 488 nm, and detection windows were at 514 to 526 nm (donor)
trypsinized and seeded at a density of 2 × 105 cells per substrate. and 566 to 578 nm (acceptor), with a 50/50 beam splitter dividing the
The medium for experiments was additionally supplemented with as- emitted light between the two detectors. All settings (photomultiplier tube
corbic acid (50 mg/ml) to stimulate collagen assembly, unless indicated voltage, pixel dwell time, and laser power) were kept constant across all
otherwise. To down- or up-regulate cell contractility, we supplemen- FRET experiments. EdU and other stainings were imaged on a Leica TCS
ted the medium with blebbistatin (10 mM, Sigma-Aldrich) or human SMD confocal laser scanning microscope using a 20× air or a 63× water-
recombinant TGF-b1 (1 ng/ml, Invitrogen), respectively. After seeding, immersion objective. Fluorophores were excited with 405-, 488-, 546-,
medium was exchanged every 2 to 3 days during the entire experi- and 633-nm laser lines, with detection windows set according to the emis-
ment. All cultured cells in the laboratory were routinely checked for sion band of the labels. Confocal stacks for the different channels were
mycoplasma infection. acquired in sequential mode to minimize cross-talk between the different
Fixation and immunofluorescence acquisition channels. Only scaffold corners where tissue had formed were
After the end of the experiment, the scaffolds were fixed in 3% para- included in the analysis.
formaldehyde (Thermo Fisher Scientific) for 15 min and permeabilized Detection of tissue boundaries and 3D age-distance map
in 1% Triton X-100 (Sigma-Aldrich) for 20 min. The actin cytoskeleton All image and data analyses were carried out using custom-written
and cell nuclei were stained with Alexa Fluor 488 phalloidin (1:1000, MATLAB scripts (MATLAB R2016a, MathWorks) on a Xeon E5-2697
A12379, Thermo Fisher Scientific) and DAPI (2 mg/ml, D1306, Invitrogen), v2 machine with 12 cores at 2.7 GHz and 64 GB RAM (random-access
respectively. Col-I primary antibody (ab90395, Abcam plc) was added memory). Confocal image stacks were converted from vendor-specific
to the medium at a dilution of 1:200 30 min before fixation to facilitate formats into MATLAB arrays using the OME Bio-Formats library (45).
ECM staining, as described previously (44), and counterstained with Tissue orientation was determined automatically from the intensity
Alexa Fluor 633 goat anti-mouse secondary antibody (A-21052, Ther- distribution, and stacks were rotated around the z axis such that the
mo Fisher Scientific) at 1:50. YAP was stained with YAP (63.7) primary scaffold was always in the top left corner. Constant background was re-
mouse monoclonal antibody (sc-101199, Santa Cruz Biotechnology) at moved by finding the first maximum of the intensity histogram, sub-
1:200 and counterstained with Alexa Fluor 633 goat anti-mouse second- tracting this value from all voxels, and setting negative intensities to
ary antibody (A-21052, Thermo Fisher Scientific) at 1:50. a-SMA was zero. The original image slices were then downscaled in the xy plane
stained with primary mouse anti–a-SMA antibody [0.N.5] (ab18147, to obtain isotropic 3D resolution. Scaffold boundaries were detected

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 8 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE

by summing up the downscaled stack in z direction, applying a Gaussian Image stacks were background-corrected, and individual slices were
filter (size = 7, s = 3) to the result, and finding the tissue boundaries low-pass–filtered by convolution with a Gaussian (size = 5, s = 1) to
on the top and left from the projected intensities, defined as the point reduce high-frequency noise. To obtain a mask of the nuclei, the DAPI
were 20% of the maximum intensity is reached. To detect the tissue signal was separated from the background by automated thresholding
surface, the normalized cumulative sum of the stack in the z direction using Otsu’s algorithm on each individual slice. The same thresholding
was Gaussian-filtered (size = 11, s = 5), converted to a mask using a method was applied to the images of YAP-stained tissue to obtain a
fixed threshold of 0.12, and combined with the previously detected scaf- mask of the cells, because YAP is located both in the nucleus and in
fold walls. Any remaining holes inside the tissue volume were removed the cytoplasm. Three-dimensional distance maps of the tissue from
by morphological closing. To obtain the age-distance map, the 3D the surface and from the wall were determined automatically as de-
Euclidean distance map of the tissue mask from the tissue surface was scribed above. Each voxel in the YAP channel was categorized as nuclear,
calculated, thus assigning a “tissue depth” with respect to the growth cytoplasmic, or background depending on whether it matched both
front to each location inside the tissue. This distance map was then masks (nuclei and cells), only the cell mask, or neither of the two. To
divided into bins with a spacing corresponding to the z step size. Finally, quantify the nuclear-to-cytoplasmic ratio, the YAP signal densities were
the scaffold and tissue masks, as well as the distance map, were rescaled divided by each other for each distance bin.
to the original image resolution and saved for later use. Because differ- EdU imaging
ences in tissue volume are difficult to assess if the tissue is larger than the To map cell proliferation as a function of location and tissue age, EdU
field of view, we estimated tissue size by measuring the tissue radius, that was incorporated, and nuclei were stained for proliferating cells (EdU
is, the distance between the corner of the scaffold and the growth front, positive) and counterstained using Hoechst (all cells). Confocal image

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

at a constant z depth of 40 mm for all image stacks. stacks were background-corrected, and individual slices were low-pass–
Determination of FN-FRET ratios in microtissues filtered by convolution with a Gaussian (size = 5, s = 1) to reduce high-
FRET experiments were conducted and analyzed by adapting previously frequency noise. To obtain a mask of the nuclei, both channels were
published protocols (24, 32) for our 3D microtissues. In brief, donor and individually separated from the background by automated thresholding
acceptor stacks were corrected for detector dark current, sensitivity, and using Otsu’s algorithm on each individual slice. Three-dimensional
donor bleed-through into the acceptor channel as previously described distance maps of the tissue from the surface and from the wall were
(32) before calculating the FRET intensity ratio (IA/ID). In addition, we determined automatically as described above, but using slightly
determined the chromatic aberration–induced offset of both channels in larger Gaussian kernels to account for the more localized intensity
the z direction by incorporating 100-nm TetraSpeck beads (T7279, distribution, because only nuclei were labeled in this case. To quan-
Thermo Fisher Scientific) into unstained tissues before fixation. After tify the percentage of proliferating cells, the total volume of all EdU-
applying a corresponding subpixel shift to the donor stack to align both positive nuclei was divided by the total volume of all nuclei for each
channels, the “control” FRET ratio for TetraSpeck beads remained con- distance bin.
stant throughout the tissue volume. This offset correction was only per-
formed for FRET measurements, because the other experiments did not Statistical analysis
involve pixel-wise intensity ratios and were therefore not sensitive to All parameters were tested for differences between the three conditions
subpixel offsets. Three-dimensional distance maps of the tissue from using one-way analysis of variance (ANOVA) and corrected for mul-
the surface and from the wall were determined automatically as de- tiple testing using Bonferroni correction. For the data in Fig. 2F, non-
scribed above, and resulting FRET values were correlated to distance parametric ANOVA (Kruskal-Wallis) was used. Each reported assay
and tissue age on a per-voxel basis. includes data from 8 different tissues (9 in the case of FRET) per con-
a-SMA expression dition, as stated in the figure legends, from at least two different inde-
To determine the relative amount of a-SMA expression in different loca- pendent tissue growth experiments for each condition and assay,
tions and between conditions, tissues were fixed and stained against amounting to a total of 99 tissues analyzed. In the figures, significance
a-SMA and F-actin, and confocal Z stacks of both channels were re- levels for differences between groups are indicated as *P < 0.05, **P <
corded. Image stacks were background-corrected, and individual slices 0.01, and ***P < 0.001. All data analysis was carried out using fully auto-
were low-pass–filtered by convolution with a Gaussian (size = 5, s = 1) mated scripts to prevent investigator bias.
to reduce high-frequency noise. The actin channel was automatically
thresholded using Otsu’s algorithm on individual normalized slices.
Three-dimensional distance maps of the tissue from the surface and from
the wall were determined automatically as described above. The binarized SUPPLEMENTARY MATERIALS
actin image was applied to the a-SMA stack channel as a mask to dis- Supplementary material for this article is available at http://advances.sciencemag.org/cgi/
content/full/4/1/eaao4881/DC1
criminate signal (colocalized with F-actin fibers) and background. Any
fig. S1. Individual channels of images from Fig. 1, F to H.
a-SMA signal not colocalized with F-actin fibers was treated as back- fig. S2. Individual channels of images from Fig. 2, A to C.
ground, because a-SMA is expected to localize at actin stress fibers rather fig. S3. Individual channels of images from Fig. 4, A to C.
than in the cytoplasm. The expression of a-SMA was quantified as the fig. S4. Individual channels of images from Fig. 4, G to I.
actin-colocalized a-SMA signal density (signal per actin volume) divided fig. S5. Illustration of the scaffold fabrication method.
by the background a-SMA signal density (signal per background volume) fig. S6. Gradients of cell phenotype and FN stretch after 11 days.
movie S1. Z stack showing actin (green) and nuclei (blue).
to avoid effects of cell density and actin signal strength. The results were
movie S2. Z stack showing nuclei (red) and EdU (green).
plotted as a function of distance to the surface. movie S3. Z stack showing color-coded FN-FRET signals.
YAP nuclear translocation movie S4. Z stack showing actin (green) and a-SMA (red).
To quantify the nuclear localization of YAP, tissues were fixed and movie S5. Z stack showing nuclei (red) and YAP (cyan).
stained against YAP using an antibody and against nuclei using DAPI. movie S6. Time lapse recording of a tissue with fluorescently labeled FN.

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 9 of 10

SCIENCE ADVANCES | RESEARCH ARTICLE

REFERENCES AND NOTES 28. T. Panciera, L. Azzolin, M. Cordenonsi, S. Piccolo, Mechanobiology of YAP and TAZ in
1. V. S. LeBleu, G. Taduri, J. O’Connell, Y. Teng, V. G. Cooke, C. Woda, H. Sugimoto, R. Kalluri, physiology and disease. Nat. Rev. Mol. Cell Biol. 18, 758–770 (2017).
Origin and function of myofibroblasts in kidney fibrosis. Nat. Med. 19, 1047–1053 (2013). 29. T. Yamashita, P. Kollmannsberger, K. Mawatari, T. Kitamori, V. Vogel, Cell sheet mechanics:
2. B. Hinz, The role of myofibroblasts in wound healing. Curr. Res. Transl. Med. 64, 171–177 How geometrical constraints induce the detachment of cell sheets from concave
(2016). surfaces. Acta Biomater. 45, 85–97 (2016).
3. K. Kis, X. Liu, J. S. Hagood, Myofibroblast differentiation and survival in fibrotic disease. 30. K. E. Kubow, R. Vukmirovic, L. Zhe, E. Klotzsch, M. L. Smith, D. Gourdon, S. Luna, V. Vogel,
Expert Rev. Mol. Med. 13, e27 (2011). Mechanical forces regulate the interactions of fibronectin and collagen I in
4. A. B. Kobb, T. Zulueta-Coarasa, R. Fernandez-Gonzalez, Tension regulates myosin extracellular matrix. Nat. Commun. 6, 8026 (2015).
dynamics during Drosophila embryonic wound repair. J. Cell Sci. 130, 689–696 (2017). 31. W. R. Legant, A. Pathak, M. T. Yang, V. S. Deshpande, R. M. McMeeking, C. S. Chen,
5. J. W. O’Connor, E. W. Gomez, Biomechanics of TGFb-induced epithelial-mesenchymal Microfabricated tissue gauges to measure and manipulate forces from 3D microtissues.
transition: Implications for fibrosis and cancer. Clin. Transl. Med. 3, 23 (2014). Proc. Natl. Acad. Sci. U.S.A. 106, 10097–10102 (2009).
6. L. Van De Water, S. Varney, J. J. Tomasek, Mechanoregulation of the myofibroblast in 32. W. R. Legant, C. S. Chen, V. Vogel, Force-induced fibronectin assembly and matrix
wound contraction, scarring, and fibrosis: Opportunities for new therapeutic intervention. remodeling in a 3D microtissue model of tissue morphogenesis. Integr. Biol. (Camb.) 4,
Adv. Wound Care (New Rochelle) 2, 122–141 (2013). 1164–1174 (2012).
7. F. Guo, D. E. Carter, A. Leask, miR-218 regulates focal adhesion kinase–dependent TGFb 33. M. S. Sakar, J. Eyckmans, R. Pieters, D. Eberli, B. J. Nelson, C. S. Chen, Cellular forces and
signaling in fibroblasts. Mol. Biol. Cell 25, 1151–1158 (2014). matrix assembly coordinate fibrous tissue repair. Nat. Commun. 7, 11036 (2016).
8. B. Hinz, S. H. Phan, V. J. Thannickal, M. Prunotto, A. Desmoulière, J. Varga, O. De Wever, 34. P. Campinho, M. Behrndt, J. Ranft, T. Risler, N. Minc, C.-P. Heisenberg, Tension-oriented
M. Mareel, G. Gabbiani, Recent developments in myofibroblast biology: Paradigms for cell divisions limit anisotropic tissue tension in epithelial spreading during zebrafish
connective tissue remodeling. Am. J. Pathol. 180, 1340–1355 (2012). epiboly. Nat. Cell Biol. 15, 1405–1414 (2013).
9. F. Klingberg, M. L. Chow, A. Koehler, S. Boo, L. Buscemi, T. M. Quinn, M. Costell, 35. S. Lepreux, A. Desmoulière, Human liver myofibroblasts during development and
B. A. Alman, E. Genot, B. Hinz, Prestress in the extracellular matrix sensitizes latent diseases with a focus on portal (myo)fibroblasts. Front. Physiol. 6, 173 (2015).
36. M. Junkin, S. L. Leung, S. Whitman, C. C. Gregorio, P. K. Wong, Cellular self-organization by

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

TGF-b1 for activation. J. Cell Biol. 207, 283–297 (2014).
10. T. Kisseleva, M. Cong, Y. Paik, D. Scholten, C. Jiang, C. Benner, K. Iwaisako, T. Moore-Morris, autocatalytic alignment feedback. J. Cell Sci. 124, 4213–4220 (2011).
B. Scott, H. Tsukamoto, S. M. Evans, W. Dillmann, C. K. Glass, D. A. Brenner, 37. S. Guo, L. A. DiPietro, Factors affecting wound healing. J. Dent. Res. 89, 219–229 (2010).
Myofibroblasts revert to an inactive phenotype during regression of liver fibrosis. 38. A. Biernacka, M. Dobaczewski, N. G. Frangogiannis, TGF-b signaling in fibrosis.
Proc. Natl. Acad. Sci. U.S.A. 109, 9448–9453 (2012). Growth Factors 29, 196–202 (2011).
11. C. Huang, R. Ogawa, Fibroproliferative disorders and their mechanobiology. 39. A. Cipitria, C. Lange, H. Schell, W. Wagermaier, J. C. Reichert, D. W. Hutmacher, P. Fratzl,
Connect. Tissue Res. 53, 187–196 (2012). G. N. Duda, Porous scaffold architecture guides tissue formation. J. Bone Miner. Res.
12. P. Kong, P. Christia, N. G. Frangogiannis, The pathogenesis of cardiac fibrosis. 27, 1275–1288 (2012).
Cell. Mol. Life Sci. 71, 549–574 (2014). 40. C. J. Chan, C.-P. Heisenberg, T. Hiiragi, Coordination of morphogenesis and cell-fate
13. N. Nakagawa, J. S. Duffield, Myofibroblasts in fibrotic kidneys. Curr. Pathobiol. Rep. 1, specification in development. Curr. Biol. 27, R1024–R1035 (2017).
189–198 (2013). 41. J. C. Love, D. B. Wolfe, H. O. Jacobs, G. M. Whitesides, Microscope projection
14. K. E. Kubow, E. Klotzsch, M. L. Smith, D. Gourdon, W. C. Little, V. Vogel, Crosslinking of photolithography for rapid prototyping of masters with micron-scale features for use in
cell-derived 3D scaffolds up-regulates the stretching and unfolding of new soft lithography. Langmuir 17, 6005–6012 (2001).
extracellular matrix assembled by reseeded cells. Integr. Biol. (Camb.) 1, 635–648 42. P. Kim, H. E. Jeong, A. Khademhosseini, K. Y. Suh, Fabrication of non-biofouling
(2009). polyethylene glycol micro- and nanochannels by ultraviolet-assisted irreversible sealing.
15. D. E. Ingber, Reverse engineering human pathophysiology with organs-on-chips. Cell Lab Chip 6, 1432–1437 (2006).
164, 1105–1109 (2016). 43. B. Trappmann, J. E. Gautrot, J. T. Connelly, D. G. T. Strange, Y. Li, M. L. Oyen, M. A. C. Stuart,
16. J.-M. Nam, Y. Onodera, M. J. Bissell, C. C. Park, Breast cancer cells in three-dimensional H. Boehm, B. Li, V. Vogel, J. P. Spatz, F. M. Watt, W. T. S. Huck, Extracellular-matrix
culture display an enhanced radioresponse after coordinate targeting of integrin tethering regulates stem-cell fate. Nat. Mater. 11, 642–649 (2012).
a5b1 and fibronectin. Cancer Res. 70, 5238–5248 (2010). 44. T. O. Ihalainen, L. Aires, F. A. Herzog, R. Schwartlander, J. Moeller, V. Vogel, Differential
17. C. X. Li, N. P. Talele, S. Boo, A. Koehler, E. Knee-Walden, J. L. Balestrini, P. Speight, A. Kapus, basal-to-apical accessibility of lamin A/C epitopes in the nuclear lamina regulated by
B. Hinz, MicroRNA-21 preserves the fibrotic mechanical memory of mesenchymal changes in cytoskeletal tension. Nat. Mater. 14, 1252–1261 (2015).
stem cells. Nat. Mater. 16, 379−389 (2017). 45. M. Linkert, C. T. Rueden, C. Allan, J.-M. Burel, W. Moore, A. Patterson, B. Loranger, J. Moore,
18. C. M. Bidan, K. P. Kommareddy, M. Rumpler, P. Kollmannsberger, Y. J. M. Bréchet, P. Fratzl, C. Neves, D. MacDonald, A. Tarkowska, C. Sticco, E. Hill, M. Rossner, K. W. Eliceiri,
J. W. C. Dunlop, How linear tension converts to curvature: Geometric control of bone J. R. Swedlow, Metadata matters: Access to image data in the real world. J. Cell Biol. 189,
tissue growth. PLOS ONE 7, e36336 (2012). 777–782 (2010).
19. C. M. Bidan, P. Kollmannsberger, V. Gering, S. Ehrig, P. Joly, A. Petersen, V. Vogel, P. Fratzl,
J. W. C. Dunlop, Gradual conversion of cellular stress patterns into pre-stressed matrix Acknowledgments
architecture during in vitro tissue growth. J. R. Soc. Interface 13, 20160136 (2016). Funding: Parts of the research leading to these results have received funding from the
20. M. Herklotz, M. C. Prewitz, C. M. Bidan, J. W. C. Dunlop, P. Fratzl, C. Werner, Availability of European Union Seventh Framework Programme (FP7/2007–2013) under grant agreement
extracellular matrix biopolymers and differentiation state of human mesenchymal no. 327065 (P.K.), a European Research Council (ERC) Advanced Grant (233157) from
stem cells determine tissue-like growth in vitro. Biomaterials 60, 121–129 (2015). the ERC (V.V.), the Swiss Science Foundation through project CR32I3-156931 (V.V.), and
21. S. Cho, J. Irianto, D. E. Discher, Mechanosensing by the nucleus: From pathways to scaling from ETH Zurich. Author contributions: P.K., C.M.B., J.W.C.D., P.F., and V.V. designed
relationships. J. Cell Biol. 216, 305–315 (2017). research. P.K., C.M.B., and J.W.C.D. developed experimental methods and data analysis.
22. K. N. Dahl, A. J. S. Ribeiro, J. Lammerding, Nuclear shape, mechanics, and P.K. conducted experiments and analyzed data. All authors discussed the results. P.K. and
mechanotransduction. Circ. Res. 102, 1307–1318 (2008). V.V. wrote the manuscript. Competing interests: The authors declare that they have no
23. G. Baneyx, L. Baugh, V. Vogel, Coexisting conformations of fibronectin in cell culture competing interests. Data and materials availability: All data needed to evaluate the
imaged using fluorescence resonance energy transfer. Proc. Natl. Acad. Sci. U.S.A. 98, conclusions in the paper are present in the paper and/or the Supplementary Materials.
14464–14468 (2001). Additional data related to this paper may be requested from the authors. All data and the
24. M. L. Smith, D. Gourdon, W. C. Little, K. E. Kubow, R. A. Eguiluz, S. Luna-Morris, V. Vogel, MATLAB code to produce the plots from the data are available for download from the
Force-induced unfolding of fibronectin in the extracellular matrix of living cells. author homepage.
PLOS Biol. 5, e268 (2007).
25. G. Baneyx, L. Baugh, V. Vogel, Fibronectin extension and unfolding within cell matrix
fibrils controlled by cytoskeletal tension. Proc. Natl. Acad. Sci. U.S.A. 99, 5139–5143 (2002). Submitted 26 July 2017
26. W. C. Little, M. L. Smith, U. Ebneter, V. Vogel, Assay to mechanically tune and optically Accepted 11 December 2017
probe fibrillar fibronectin conformations from fully relaxed to breakage. Matrix Biol. 27, Published 17 January 2018
451–461 (2008). 10.1126/sciadv.aao4881
27. G. Nardone, J. Oliver-De La Cruz, J. Vrbsky, C. Martini, J. Pribyl, P. Skládal, M. Pešl,
G. Caluori, S. Pagliari, F. Martino, Z. Maceckova, M. Hajduch, A. Sanz-Garcia, N. M. Pugno, Citation: P. Kollmannsberger, C. M. Bidan, J. W. C. Dunlop, P. Fratzl, V. Vogel, Tensile forces
G. B. Stokin, G. Forte, YAP regulates cell mechanics by controlling focal adhesion drive a reversible fibroblast-to-myofibroblast transition during tissue growth in engineered
assembly. Nat. Commun. 8, 15321 (2017). clefts. Sci. Adv. 4, eaao4881 (2018).

Kollmannsberger et al., Sci. Adv. 2018; 4 : eaao4881 17 January 2018 10 of 10

Tensile forces drive a reversible fibroblast-to-myofibroblast transition during tissue growth in
engineered clefts
Philip Kollmannsberger, Cécile M. Bidan, John W. C. Dunlop, Peter Fratzl and Viola Vogel

Sci Adv 4 (1), eaao4881.

DOI: 10.1126/sciadv.aao4881

Downloaded from http://advances.sciencemag.org/ on January 18, 2018

ARTICLE TOOLS http://advances.sciencemag.org/content/4/1/eaao4881

SUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2018/01/12/4.1.eaao4881.DC1
MATERIALS

REFERENCES This article cites 45 articles, 13 of which you can access for free
http://advances.sciencemag.org/content/4/1/eaao4881#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science Advances (ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 New
York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive licensee American
Association for the Advancement of Science. No claim to original U.S. Government Works. The title Science Advances is a
registered trademark of AAAS.